Hardware
How NVIDIA DSX MaxLPS Maximizes AI Factory Throughput and Efficiency
NVIDIA’s DSX MaxLPS employs policy‑governed power sharing to dynamically reallocate power across GPU resources, enabling up to 40 % more GPUs within the same approved power budget. In a joint test at
NVIDIA’s DSX MaxLPS employs policy‑governed power sharing to dynamically reallocate power across GPU resources, enabling up to 40 % more GPUs within the same approved power budget. In a joint test at Nscale’s Verne campus on GB300 NVL72 systems, 192 GPUs were run under the same 264.4 kW budget that previously supported 140 GPUs, boosting normalized aggregate throughput by 49.2 % and tokens per watt from 4.10 to 6.12 while keeping per‑instance throughput and median/P75 latency within 5 % of baseline (P99 first‑token time rose 17 %). Operators validate deployment through a five‑stage process—defining the managed boundary, establishing baselines, introducing conservative policies, incrementally adding capacity with testing, and setting production limits once objectives are met—and can review NVIDIA DSX MaxLPS documentation for detailed implementation guidance.
Related
- Maximizing AI Factory Performance per Watt with NVIDIA DSX MaxLPS
- How NVIDIA Groq 3 LPX Deterministic Execution Drives Power-Efficient High-Interactivity Inference on NVIDIA Vera Rubin
- Scaling AI Inference Across Multiple GPUs Using NVIDIA TensorRT with Multi-Device Inference Support
- Simplifying Model Serving Across Multiple GPUs with NVIDIA TensorRT Multi-Device Integration in NVIDIA Dynamo-Triton
Source: NVIDIA Developer | 2026-09-28